Tracking Issue for Generic Constant Arguments MVP

[camelid]

This is a tracking issue for the prototype described in rust-lang/rust-project-goals#100. Background reading as to the design/justification for this feature.
The feature gate for the issue is #![feature(min_generic_const_args)].

About tracking issues

Tracking issues are used to record the overall progress of implementation.
They are also used as hubs connecting to other relevant issues, e.g., bugs or open design questions.
A tracking issue is however not meant for large scale discussion, questions, or bug reports about a feature.
Instead, open a dedicated issue for the specific matter and add the relevant feature gate label.
Discussion comments will get marked as off-topic or deleted.
Repeated discussions on the tracking issue may lead to the tracking issue getting locked.

Steps

• Add ConstArgKind::Path (Represent type-level consts with new-and-improved hir::ConstArg #125915)
• Use ConstArgKind::Path for all single-segment paths (Use ConstArgKind::Path for all single-segment paths, not just params under min_generic_const_args #131081)
• Use ConstArgKind::Path for all paths
  • most paths mgca: Lower all const paths as ConstArgKind::Path #135186
  • Support ConstArgKind::Path for uses of static paths #132986
• Support "simple" exprs necessary for constructing adt_const_params types:
  • Struct construction exprs, e.g. Adt { field: N }
    • Core impl MGCA: Support struct expressions without intermediary anon consts #149114
    • Rest patterns MGCA: support struct expressions with ..rest patterns as direct const args? #150622
  • Tuple constructor calls, e.g. Some(N) MGCA: support tuple constructor calls as a direct const argument #150610
  • Const constructs, e.g. None Support ConstArgKind::Paths for const struct/variant constructors #132985
  • Array exprs, e.g. [M; N], [1, 2, N] MGCA: Support array expressions as direct const arguments #150612
  • Tuple exprs, (10, N) MGCA: Support tuple expressions as direct const arguments #150613
• reject impls not using #[type_const] on their assoc consts if the trait used it mgca: Finish implementation of #[type_const] #148716
• enforce type_const marked const items have a type that implements ConstParamTy mgca: Finish implementation of #[type_const] #148716
• forbid generic parameters from being used in the types of const parameters MGCA: forbid generic parameters from being in used in the types of type_const items #150614
• lower uses of type_consts to mir::Const::Ty so they can be normalized in the MIR correctly MGCA: uses of type_const items inside of bodies should lower to mir::Const::Ty #150615
• Make inherent associated constants behave correctly rustc internal compiler error: 'rustc' failed at 'Box<Any>' #15200
• Implement type system normalization that makes use of associated_const_equality bounds mgca: Add ConstArg representation for const items #139558
• Merge feature(associated_const_equality) into min_generic_const_args MGCA: merge associated_const_equality feature gate into MGCA #150617
• Experiment with adding some kind of ConstArgKind::Literal MGCA: Support literals as direct const arguments #150618
• Traits are dyn compatible if all associated consts are type safe Permit trait object types where all (non-generic) associated constants are specified (via assoc item bindings) #130300
• Find a way to avoid code duplication between HIR ty lowering and typechecking for struct expressions MGCA: Avoid code duplication between HIR ty lowering and elsewhere #150621
• Make HIR ty lowering lower uses of generic parameters to errors if in an item with no generics
  • ish, Use more principled check for generics in const ops #150519
  • Validate HIR ty lowering's produced generic parameters #150620
• Find better names for ConstKind::Value and ValTree now that they're not fully evaluated Followup work from making ValTree recurse through ty::Const #150624
• Do something about CTFE returning ValTrees with erased lifetimes. Likely means adding a ValTree which doesn't recurse through ty::Const like how it used to be Followup work from making ValTree recurse through ty::Const #150624
• Typecheck that const arguments outside of paths have the correct type MGCA: Check that const arguments in struct exprs are correctly typed #150623
• Adjust documentation (see instructions on rustc-dev-guide)
• Stabilization PR (see instructions on rustc-dev-guide)

Unresolved Questions

• Can we get rid of needing an annotation on const items for them being type consts?
  • If no we need t-style to decide on rustfmt stuff

Implementation history

Implementation history

• Initial document proposing the design/feature hackmd link
• Represent type-level consts with new-and-improved hir::ConstArg #125915
• Fix anon const def-creation when macros are involved #129137
• Use ConstArgKind::Path for all single-segment paths, not just params under min_generic_const_args #131081
• Encode MIR for const ctors in metadata and allow eval for them #134873
• mgca: Lower all const paths as ConstArgKind::Path #135186
• Set groundwork for proper const normalization #140549
• mgca: Add ConstArg representation for const items #139558
• mgca: Finish implementation of #[type_const] #148716
• MGCA: Syntactically distinguish anon const const args #149136
• dont create unnecessary DefIds under mgca #150025
• MGCA: Support struct expressions without intermediary anon consts #149114
• avoid recreating field_const vec #150380
• refactor destructure_const #150411
• Use more principled check for generics in const ops #150519
• Remove unneeded forbid_generic field from Res::SelfTyAlias #150589
• MGCA: Support for tuple constructors  #150603
• Forbid generic parameters in types of #[type_const] items #150650
• MGCA: Support tuple expressions as direct const arguments #150675
• Merge associated_const_equality feature gate into MGCA #150640
  • This tracking issue now inherits the impl history of ACE's tracking issue Tracking issue for associated const equality #92827
  • allow eq constraints on associated constants #87648¹
  • Continue work on associated const equality #93285²
  • rustdoc: handle assoc const equalities in cross-crate APIT #113028
  • Resolve associated item bindings by namespace #118668
  • Fix type resolution of associated const equality bounds (take 2) #119385
  • Reject overly generic assoc const binding types #121258
  • rustdoc: Support middle::ty associated const equality predicates again #142092
  • mGCA: Move tests for assoc const bindings (formerly ACE) into dedicated directory & replace more mentions of ACE #150844
  • Support v0 demangling of trait object types with assoc const bindings rustc-demangle#87
  • mGCA: Make trait object types with type-level associated consts dyn compatible #150843
• MGCA: Support literals as direct const arguments #150699
• MGCA: Const constructors support #150704
• MGCA: pretty printing for struct expressions and tuple calls  #150695
• mgca: Type-check fields of struct expr const args #150713
• Fix ICE: can't type-check body of DefId for issue #148729 #150799
• Fix ICE: When Trying to check visibility of a #[type_const], check RHS instead. #151085
• MGCA: Fix incorrect pretty printing of valtree arrays #151296
• Added mGCA related tests #150886
• Emit error instead of delayed bug when meeting mismatch type for const array #151055
• mGCA: Support array expression as direct const arguments #150786
• Use find_attr instead of attr::contains_name in lower_const_item_rhs #151288
• Normalize type_const items even with feature generic_const_exprs #151275
• Remove FeedConstTy and provide ty when lowering const arg #150962
• Emit error instead of delayed bug when meeting mismatch type for const tuple #150869
• Fixed ICE when using function pointer as const generic parameter #151219
• Include assoc const projections in CFI trait object #151947
• Check proj's parent is trait or not when checking dyn compatibility #151754
• Fix ICE in normalizing inherent associated consts with #[type_const] #152000
• mGCA: Validate const literal against expected type #152001
• Do not emit ConstEvaluatable goals if type-const #152040
• MGCA: require #[type_const] on free consts too #152129
• Consolidate type const checks on tcx.is_type_const #152136
• mGCA: Support directly represented negated literals #152139
• mGCA: Add associated const type check #152146
• Fix const normalization for generic const items with trait assoc consts #152278
• Update mgca to use type const syntax instead of the #[type_const] attribute.  #152324
• Feature gate for defaulted associated type_consts with associated_type_defaults  #152385
• mGCA: Enforce WF element types for array valtrees #152492 mgca: Type-check field of array expr const args #152458
• fix missleading error for tuple ctor #152565
• Fix ICE in try_to_raw_bytes when array elements have mismatched #152794
• Make const_lit_matches_ty check literal suffixes for exact type match #152906
• mGCA: improve ogca diagnostic message  #152952
• refactor(mgca): Change DefKind::Const and DefKind::AssocConst to have a is_type_const flag #153050
• mGCA: Lower negated literals directly and reject non-integer negations #153075

Footnotes

1. Initial effort to implement which includes parsing and use of term throughout the codebase, but still lacking a complete implementation. ↩

2. More thorough implementation which works for basic cases. This allowed for consts to actually be bound, more than just parsing them. ↩